The second group's philosophy was to give users as much power to focus on their specific interests as possible. The group decided that a necessary precondition for such directednesss would be a uniform and intuitive organization, but to truly put the power in the hands of users required giving them options. This was done by providing three access methods: elaborate indexing, a graphical overview, and a searchable index.
The final organization of the Group 2 document consisted of a broad, shallow organization in which topic nodes were accessible from the the introductory page and a specific discussion of the topic in relation to an architecture was then accessible from the topic pages. This is outlined in Figure 2. Individual topics were made nodes in the Group 2 document with many intra-document links connecting ideas within a topic. Such links are represented in Figure 2 by the links within the ARCHX node. Both topic and architecture-specific nodes referenced nodes inside and outside the current node where appropriate. Thus, outside of the top level's modularity, the document's organization was more fluid than the others, depending primarily on the interconnectivity of the linked topics for navigation.
To help the user navigate, the Group 2 graphical overview was made extremely extensive, providing an intuitive feel for the organization of the document as a whole. All nodes and most sub-topics within a node were directly accessible from the overview. All the information was made clickable, so users could either get a high-level discussion of a topic or zoom in on specifics directly from the graphic map.
Finally, Group 2 built a case-insensitive search engine, placed on the introductory page for easy access, to allow the user immediate access to any keywords in the document. The searcher searched the whole document, creating a file that had links to the matched files. The search engine gave users ultimate indexing ability: providing them with a means to comb the whole document for information on their specific interests in a manner the documents authors' could not have foreseen. This search engine was based on a high-level (``C'' language) interface between the Mosaic server and several UNIX commands for searching and file creation. This interface did not use knowledge of Mosaic to find appropriate text but rather searched through a specified directory tree of files to find appropriate text and then generated an HTML document to access those files via Mosaic.
